Introduction

Overview and Motivation

In political literature, democracy is considered the most reliable and fair regime in the eyes of citizens. There are several forms of democracy in our World and they all adapt to the country’s socio-demographic, physiologic and cultural values. For this reason, it is impossible to give a unique definition of what the word “democracy” means. However, what we know is that it is a form of government in which the people have the authority to choose their governing legislation.

The opposite of this form of government is the autocratic regime. In that case, government rules do not allow people to participate directly in the governing process. The power is held by one single entity or by a minority group. We will not go into details which regime is the best for the behalf of the country, but we will focus more on how countries with more or less democracy level react to Covid-19 pandemic.

Research questions

As we all know, Covid-19 has become much more than a health problem. Fighting the pandemic, the majority of countries are facing pressure in the health sector but, at the same time, they all need to deal with their economic and socio-cultural capability under their government rules. In our work, we are interested in analysing how countries with different government regime behave during the fight against Covid-19 and how restrictions differ between those countries. We aim to evaluate the validity of the political theory stating that the democracy is the best regime to deal with the crisis.

Our research is a comparative analysis of European countries where we are going to compare restriction policies between European countries with different degrees of democracy.

More precisely, we are trying to answer the following questions:

  1. How different is the mortality rate between countries with high and low democracy level?
  2. How severe those countries are in term of movement restrictions? With focus on stay-home and public gathering restriction.
  3. What about the right to health? How testing policies and contact tracing differ between different level of democracy?
  4. Is it true that countries with the highest democracy level achieve better results, in term of people infected, through restriction?

Data

For our research, we are going to use seven different datasets.

Covid data

The first is the dataset with information related to covid. We downloaded it from the portal https://ourworldindata.org. It is composed of many information related to covid for almost all countries in the world. The available observations go from “2019-12-31” to “2020-11-07”, and we have information about:

Table 1: covid_data, list of variables availabe in our dataset.
iso_code icu_patients_per_million population
continent hosp_patients population_density
location hosp_patients_per_million median_age
date weekly_icu_admissions aged_65_older
total_cases weekly_icu_admissions_per_million aged_70_older
new_cases weekly_hosp_admissions gdp_per_capita
new_cases_smoothed weekly_hosp_admissions_per_million extreme_poverty
total_deaths total_tests cardiovasc_death_rate
new_deaths new_tests diabetes_prevalence
new_deaths_smoothed total_tests_per_thousand female_smokers
total_cases_per_million new_tests_per_thousand male_smokers
new_cases_per_million new_tests_smoothed handwashing_facilities
new_cases_smoothed_per_million new_tests_smoothed_per_thousand hospital_beds_per_thousand
total_deaths_per_million tests_per_case life_expectancy
new_deaths_per_million positive_rate human_development_index
new_deaths_smoothed_per_million tests_units
icu_patients stringency_index

Now, we do not need all of those variables For this reason, we are going to select only those we consider useful to make a comparison of the pandemic situation in terms of cases and deaths.

  • date;
  • continent;
  • location;
  • total_cases;
  • new_cases;
  • total_deaths;
  • new_deaths;
  • population;

First, the variable location indicates the name of the country. As there is a function called location () , to avoid possible errors, we prefer to rename this variable country. Secondly, in our project, we are focusing our attention on European countries, and for this reason, we want to remove all countries that are part of other continents.

Table 2: European countries.
Albania Finland Latvia Romania
Andorra France Liechtenstein Russia
Austria Germany Lithuania San Marino
Belarus Gibraltar Luxembourg Serbia
Belgium Greece Macedonia Slovakia
Bosnia and Herzegovina Guernsey Malta Slovenia
Bulgaria Hungary Moldova Spain
Croatia Iceland Monaco Sweden
Cyprus Ireland Montenegro Switzerland
Czech Republic Isle of Man Netherlands Ukraine
Denmark Italy Norway United Kingdom
Estonia Jersey Poland Vatican
Faeroe Islands Kosovo Portugal

Democracy index

The second dataset is the democracy index dataset.

Our data come from V-DEM ( Varieties of Democracy) website, https://www.v-dem.net/en/data/data/v-dem-dataset/. V-Dem Institute is part of the Department of Political Science at the University of Gothenburg, Sweden.

There are many informations included in the dataset but, one more time, we are not going to use all of them. The original file was containing 1817 variables but we needed only 5 of them. we decided to use only two indicators of democracy which are: liberal democracy index and electoral democracy index. Therefore, we reduced this dataset in excel and load in r only a part of it.

First, we will extracted the relevant information:

  • Country;
  • Date;
  • Year;
  • v2x_libdem (Liberal democracy index);
  • v2x_polyarchy (Electoral democracy index);

These democracy indexes assume a value from 0 to 1, where 0 represents the absence of democracy while 1 represents a full democracy. One of the challenging parts of our work was to choose the appropriate variables to describe the level of democracy. Our choice derives from several existing scientific literature related to government policy and Covid-19 pandemic.

Our limitation is that our data is not updated for the year 2020. We have information up to 31 December 2019, and for this reason, we have to assume that level of democracy did not change between the end of 2019 and first semester of 2020.

Table 3a: democracy dataset
country v2x_polyarchy v2x_libdem
Mexico 0.71 0.49
Suriname 0.74 0.55
Sweden 0.87 0.83
Switzerland 0.87 0.83
Ghana 0.72 0.61
a Representation of the first 5 observation of the dataset. Total observations : 179

Stringency index

We retrieved this dataset from https://ourworldindata.org/policy-responses-covid.

The stringency index indicates the country’s level of stringency in term of restrictions applied. The variables used to evaluate this information are:

  • school closures;
  • workplace closures;
  • cancellation of public events;
  • restrictions on public gatherings;
  • closures of public transport;
  • stay-at-home requirements;
  • public information campaigns;
  • restrictions on internal movements;
  • international travel controls.

The index, on any given day, is calculated as the mean score of the nine metrics, each taking a value between 0 and 100, where 0 represent no restrictions at all and 100 maximum level of stringency. The time frame for this information goes from 2020-01-01 to 2020-11-09.

We are going to use this information to understand which countries are more severe and to evaluate their policy’s consequences in term of people infected.

The dataset has 4 variables:

  • Entity;
  • Code;
  • Date;
  • Stringency Index;

We want to have one variable for each country where each observation represents one date and the value the stringency level. The country code can be excluded, and we fill the missing values at the end of the data with the previous value.

Table 4a: tail of stringency dataset
date Afghanistan Albania Algeria Andorra
2020-11-05 5.6 51 75 59
2020-11-06 5.6 51 75 59
2020-11-07 5.6 51 75 59
2020-11-08 5.6 51 75 59
2020-11-09 5.6 51 75 59
2020-11-10 5.6 51 75 59
a Representation of the last 6 observation and first 5 columns of the dataset. Total observations : 315 . Total countries : 185

Testing policy

It was downloaded from https://ourworldindata.org/policy-responses-covid.

This dataset show to what extent the testing policies are applied in different countries. The type of test considered is the PCR test which is used to directly detect the presence of an antigen, rather than the presence of the body’s immune response, or antibodies. For this reason, we assume that there is no other available test than PCR, even if we know that PCR is not the only way to trace the presence of the virus in a sample.

Similarly to the stringency index, we have four variables:

  • Entity;
  • Code;
  • Date;
  • Testing_policy;

In this case, the testing policy index is an integer between 0 and 3:

  1. No testing policy;
  2. Only those who both (a) have symptoms AND (b) meet specific criteria (e.g. key workers, admitted to hospital, came into contact with a known case, returned from overseas);
  3. Testing of anyone showing COVID-19 symptoms;
  4. Open public testing (e.g. “drive-through” testing available to asymptomatic people);

As we did before, we are going to organize our data such that we want one variable for each country and one observation for each date and all NAs replaced with the previous value.

Table 5a: tail of testing dataset
date Afghanistan Albania Algeria Andorra
2020-11-25 3 2 1 3
2020-11-26 3 2 1 3
2020-11-27 3 2 1 3
2020-11-28 3 2 1 3
2020-11-29 3 2 1 3
2020-11-30 3 2 1 3
a Representation of the last 6 observation and first 5 columns of the dataset. Total observations : 335 . Total countries : 180

Contact tracing policy

It was downloaded from https://ourworldindata.org/policy-responses-covid.

This information indicates whether a country applies policies to trace back the infection The variable here still the same as in the previous two datasets and the value indicate:

  1. No contact tracing;
  2. Limited contact tracing - not done for all cases;
  3. Comprehensive contact tracing - done for all cases;

We want to have this information in the same format as the previous one.

Table 6a: tail of contact_tracing dataset
date Afghanistan Albania Algeria Andorra
2020-11-20 1 2 2 2
2020-11-21 1 2 2 2
2020-11-22 1 2 2 2
2020-11-23 1 2 2 2
2020-11-24 1 2 2 2
2020-11-25 1 2 2 2
a Representation of the last 6 observation and first 5 columns of the dataset. Total observations : 330 . Total countries : 179

Contract tracing and testing policy are going to tell us how different countries have supported the right to health. We consider an equitable access for testing and contact tracing policy for all people. These are supposed to be the two most important rights that governments should ensure in order to be considered as democratic. For this reason, we want to analyze those elements with the level of democracy to identify the countries that are working the most in this direction and compare those results with the evolution of the infection in such country.

Stay-at-home and public gathering restriction

It was downloaded from https://ourworldindata.org/policy-responses-covid.

Another useful information to answer our questions is ’Stay-at-home and public gathering restrictions. Both of them indicate to what extent the government is strict in term of the restrictions.

Stay-at-home restriction is represented with a value between 0 and 3:

  1. No measures;
  2. Recommend not leaving the house;
  3. Require not leaving the house with exceptions for daily exercise, grocery shopping, and ‘essential’ trips;
  4. Require not leaving the house with minimal exceptions (e.g. allowed to leave only once every few days, or only one person can leave at a time);

And, public gathering restriction information goes from 0 to 4:

  1. No restrictions;
  2. Restrictions on huge gatherings (the limit is above 1000 people);
  3. Restrictions on gatherings between 100-1000 people;
  4. Restrictions on gatherings between 10-100 people;
  5. Restrictions on gatherings of less than ten people;

We are going to use this data to see how countries with different intensity of democracy have restricted the freedom of movement. To do that, we decided to organize the last two datasets in the same structure of the others: one observation for each date where each variable identify one country.

Table 7a: tail of stay_home dataset
date Afghanistan Albania Algeria Andorra
2020-11-25 0 2 2 1
2020-11-26 0 2 2 1
2020-11-27 0 2 2 1
2020-11-28 0 2 2 1
2020-11-29 0 2 2 1
2020-11-30 0 2 2 1
a Representation of the last 6 observation and first 5 columns of the dataset. Total observations : 335 . Total countries : 178
Table 8a: tail of public_gather dataset
date Afghanistan Albania Algeria Andorra
2020-11-05 0 4 4 4
2020-11-06 0 4 4 4
2020-11-07 0 4 4 4
2020-11-08 0 4 4 4
2020-11-09 0 4 4 4
2020-11-10 0 4 4 4
a Representation of the last 6 observation and first 5 columns of the dataset. Total observations : 315 . Total countries : 185

Exploratory data analysis

After loading our data, we realized that the combination of countries is not unique for all of them.

For this reason, we need to define the list of countries common for all our data. We could do that using a joint function, but instead, we prefer to have the list of countries (countries.list) we are using, so in case we need more data, we have it ready to filter the new source for the countries selected.

Table 9: list of european countries common for all our data
Albania Kosovo
Austria Latvia
Belarus Lithuania
Belgium Luxembourg
Bosnia and Herzegovina Moldova
Bulgaria Netherlands
Croatia Norway
Cyprus Poland
Czech Republic Portugal
Denmark Romania
Estonia Russia
Finland Serbia
France Slovakia
Germany Slovenia
Greece Spain
Hungary Sweden
Iceland Switzerland
Ireland Ukraine
Italy United Kingdom

We now want to show some characteristic of those countries. From the map, we can explore them and have details about population, life expectancy, GDP per capita, median age and human development index. We will do that joining the dataset Countries_lat_lon to a new dataset containing the variables of interest previously named.

We built Countries_lat_lon dataset copying the information from Dataset Publishing Language portal offered by Google: https://developers.google.com/public-data/docs/canonical/countries_csv We created then a csv file through excel, and we loaded it in r and filtered for the countries in countries.list.

Now, we want to look for possible mistakes and error in our datasets.

covid_data

#>       date              country            new_cases    
#>  Min.   :2019-12-31   Length:14325       Min.   :-1385  
#>  1st Qu.:2020-04-11   Class :character   1st Qu.:    0  
#>  Median :2020-06-20   Mode  :character   Median :   35  
#>  Mean   :2020-06-17                      Mean   :  818  
#>  3rd Qu.:2020-08-29                      3rd Qu.:  302  
#>  Max.   :2020-11-07                      Max.   :60486  
#>                                          NA's   :150    
#>   total_cases        new_deaths     total_deaths  
#>  Min.   :      1   Min.   :-1918   Min.   :    1  
#>  1st Qu.:    574   1st Qu.:    0   1st Qu.:   32  
#>  Median :   4474   Median :    0   Median :  248  
#>  Mean   :  57379   Mean   :   20   Mean   : 3661  
#>  3rd Qu.:  32523   3rd Qu.:    6   3rd Qu.: 1614  
#>  Max.   :1733440   Max.   : 2004   Max.   :48475  
#>  NA's   :1487      NA's   :150     NA's   :3110   
#>    population       
#>  Min.   :      809  
#>  1st Qu.:   628062  
#>  Median :  5421242  
#>  Mean   : 15706866  
#>  3rd Qu.: 10708982  
#>  Max.   :145934460  
#> 

The first thing we notice are the numerous NAs presents in new_cases , total_cases, new_deaths and total_deaths. The second problem we retrieve from the summary is the negative values in the variables new_cases and new_deaths. We did some research and we found out that on those dates, a new system of gathering data had allowed them to identify cases that were counted twice and exclude deaths wrongly attributed to the virus. For this reason, we assume that on those dates, the new_cases and new_deaths are equal to 0.

After replacing the negative values, we need to fix all NAs. For now, the dataset is structured in a way that we cannot identify which observation represents the last raw of a country and which one the first of the next one. See table 10. To solve this problem, we are going to separate our four variables in four different datasets, and we will treat them separately. The reason why we are doing so is that total_casesand total_deaths are cumulative values and we want to replace NAs with the previous value plus new daily cases/deaths. If we do so, with the actual structure, we will end up assigning the last value of the last observation of a country (2020-11-07) to the first observation of the next country (2020-12-31).

Table 10: Visualization of the problem we would have fixing total_casesand total_deaths from the actual dataset.
date country new_cases total_cases new_deaths total_deaths population
2020-11-05 France 40558 1543321 385 38674 65273512
2020-11-06 France 58046 1601367 363 39037 65273512
2020-11-07 France 60486 1661853 828 39865 65273512
2019-12-31 Germany 0 NA 0 NA 83783945
2020-01-01 Germany 0 NA 0 NA 83783945
2020-01-02 Germany 0 NA 0 NA 83783945

The NAs in the variables new_casesand new_deaths can be replaced with 0. Doing so, we are assuming that no cases or deaths are registered on that date. On the other hand, total_casesand total_deaths NAs seem to appear when the value does not change from the day before. For this reason, we will replace those missing values with the previous one, except for the first observation, which will be assigned the value 0.

Now our dataset is clean and ready to use.

We are going to create new variables that will be useful to evaluate the current situation in term of infection mortality.

\[Mortality\ rate= ((total\_deaths/total\_cases) * 100)\\ Cases\ per\ 100000\ people = (total\_cases/population)*100000\\ Deaths\ per\ 100000\ people = (total\_deaths/population)*100000\\ Daily\ cases\ growth = ((total\_cases_t - total\_cases_{t-1})/total\_cases_{t-1})*100\\ Daily\ deaths\ growth =((total\_deaths_t - total\_deaths_{t-1})/total\_deaths_{t-1})*100\]

One of the most harmful consequence of Covid-19 across the world is mortality rate. It indicates how many people, out of 100 people infected, die in a specific country because of the pandemic. The mortality rate can differ from one country to another. According to the scientific report “Explaining among country variation in Covid19 case fatality rate” published in “https://www.nature.com/articles/s41598-020-75848-2”, differences in mortality rate can be caused by several factors:

Mortality tends to be more frequent in countries where the number of elderly people is higher. Related to that, It is also possible to observe a positive correlation between people with a chronic respiratory disease, cancer and smoking rate in people over 70 years.

Also GDP per capital and political regime are seen as an explicative factor to mortality such that higher GDP per capital and higher democracy level have a positive relation with the mortality rate.

Some other external resources explain mortality differences among countries in different healtcare system. Countries with a high level of testing policy allows them to identify cases early and to take necessary measures for it.

#>       date              country            mort_rate   
#>  Min.   :2019-12-31   Length:11818       Min.   : 0.0  
#>  1st Qu.:2020-03-17   Class :character   1st Qu.: 0.2  
#>  Median :2020-06-03   Mode  :character   Median : 2.4  
#>  Mean   :2020-06-03                      Mean   : 3.4  
#>  3rd Qu.:2020-08-20                      3rd Qu.: 4.3  
#>  Max.   :2020-11-05                      Max.   :19.6

As we can see from the graph, the mortality rate increased in every country from march. Italy, United Kingdom, Belgium and France have shown a mortality rate much higher than the average. In contrast, we see that countries like Iceland, Belarus and Slovakia have shown an average mortality rate lover than 1. In the analysis section we will analyze the reason of such difference, to understand if there is any correlation with the democracy level and the restriction applied.

Democracy index

As the first thing, we can remove the variable date as we assumed that the democracy index is constant for the whole period. Then, we want to select only the countries related to our project. We do that filtering the dataset with our countries list.

Let’s have a look at the map to have a visualization of our data.

As it has been argued in the “Data” part, v2x_libdem represents “Liberal democracy index”, and v2x_polyarchy represents “Electoral democracy index”. From the graphs, we can identify 2 clusters of countries: the first is characterised by a liberal democracy above 0.7 and electoral democracy above 0.80, while the second cluster shows liberal democracy lower than 0.55 and the electoral one below 0.7. We can also observe that West Europe countries have a higher level of democracy compared to countries in East Europe. Scandinavian countries are the one showing the highest level of democracy. Finally, countries’ electoral democracy index is always higher than its liberal democracy index. However, this small detail is negligible since the differences of values between them are minimal.

Stringency index

The next dataset we are going to explore is the Stringency index.

Here we isolated the countries that have been showing the highest mortality rate and we see that the level of stringency sharply increased in march but their path do not differ too much from the average level in Europe.

Results for Serbia and Bealrus are quite suprising. The first is the only country who reached the maximum level of stringency (100). It stayed constant at its high level from march till the end of April, to drop then to 25 in the month of June. On the other hand, the second one, Belarus, seems to not have applied restrictions as much as other countries. Looking at the previous graph Liberal democracy index VS electoral democracy index. This mean that there might be correlation between restrictions and democracy level. We will analyze those results more in detail in the next section.

Testing policy

#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>    0.00    1.00    2.00    1.65    2.00    3.00

This graph shows countries testing strictness index for Covid-19. First of all, we observe that countries increase the strictness of their testing policy during the first wave. Now, even there is a bit decrease, we can see clearly that they keep a certain degree of that policy (with no “0” value). They are almost all in level 2, meaning that people, who demonstrate any of the Covid symptoms, is allowed to be tested. Only Bulgaria is in level 1 showing the lack of strictness compared to other countries.

An important information we want to get from this data set is: for how long each restriction level last in each country? How many time each country has changed restriction policy? To answer those question we built out own function count.leveldays

Contact tracing policy

#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>    0.00    1.00    2.00    1.31    2.00    2.00

As indicated in the “data” part, contact tracing policy is simply the process of identification of people that a Covid patient has come in contact with. We were expecting to see a high level of control almost everywhere. However, we notice that not all of them did contact tracing. In Bosnia, for example, we can see that they never did, while Slovakia shows an opposite strategy. On average, the level is moderately high even if not constant during the period.

Stay at home

#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>    0.00    0.00    1.00    0.71    1.00    3.00

This policy is probably the one applied less. We can see that the average is 0.71, that considering the scale goes from 0 to 3, it is a low average. Countries like Norway, Belarus, Iceland, Slovenia and more, have barely applied the stay home restriction. This is a piece of exciting information for our research because the order to stay home is probably the one that more goes against people freedom. We will see in the next section if there is any correlation with the level of democracy.

Public gathering

#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>     0.0     0.0     3.0     2.3     4.0     4.0

This seems to be an interesting variable to analyze, and this will be very useful in our analysis to see how countries, with different democracy level, implement restrictions rules for gatherings. However, now, we should focus on descriptive analysis. All countries are quite strict for that policy. They show a pattern around 3 and 4, while Belarus is the only country that shows no stringency level. Although, Iceland had a no stringency level for “Stay at home policy”, but for public gathering restrictions, it shows an index of 3, which is relatively high.

Analysis

Mortality rate

We belive that countries with a high level of democracy might find harder to apply restrictions rather than countries with a low level of democracy. In democracy, citizens have the right to participate, directly or indirectly, in governing the country’s rules. However, when the government impose a restriction to fight the spread of the virus, it happened, in some countries, that people felt their rights violated because the decision did not come from them (riots around Europe).

The first information we are going to analyze is the difference in mortality rate between countries with high democracy level and low democracy level in order to identify which countries are the most affected by the pandemic.

The democracy index used in this work is “Liberal Democracy Index” since this variable seems to be the most appropriate one for our project. It measures the civil liberties and minority rights against the “tyranny of the state”. In an environment where strict rules are implemented to restrict people’s liberties, we consider this index, rather than the electoral one, more accurate.

Looking at the democracy graph in the previous section, we identified two clusters. We will name them group 1 for countries with a high level of democracy (greater than 0.69), and group 2 those with low level of democracy (lower than 0.55). We have 65% of observation in Group 1 and 35% in group 2. To keep our visuals more precise, we are going to focus on the ten countries with the highest and the ten with the lowest index.

Table 11: Countries with highest and countries with lowest level of electoral democracy.
Top Countries TDem. index Low Countries LDem. index
Denmark 0.86 Russia 0.10
Estonia 0.84 Belarus 0.12
Sweden 0.83 Serbia 0.25
Switzerland 0.83 Ukraine 0.29
Norway 0.82 Bosnia and Herzegovina 0.34
Belgium 0.82 Hungary 0.40
Portugal 0.82 Kosovo 0.41
Spain 0.81 Albania 0.43
Finland 0.81 Romania 0.43
Ireland 0.80 Moldova 0.44

Now, we are going to visualize the mortality rate, for the last available date “2020-11-05” in those two groups of countries, to see if we can observe any correlation between mortality rate and democracy level.

Average mortality rate in countries with highest and countries with lowest level of electoral democracy.
High Democracy Low Democracy
2.2 2.2

Our results are not giving a clear answer about the correlation between the two variables.
As we can see from the average mortality rate, the two groups are showing similar results. In both cases we found that at least 50% of the countries considered are suffering two deaths every 100 people infected. Sweden is the only country showing a rate above 4%, followed by Kosovo and Irland, where the fatality is slightly more than 3%.

Overall, we did not find any pattern suggesting a link between mortality rate and level of democracy. So, as we were not entirely convinced with those results, we tried to reformulate our question.

How democracy level change between countries with high and low mortality rate?

To answer the previous question, we selected the countries with the highest and the lowest democracy level,now, we are going to proceed the same way as for mortality rate.

Table 12: Countries with highest and lowest level of mortality rate.
Top Countries Mortality Rate Low Countries Mortality Rate
Italy 5.0 Iceland 0.34
United Kingdom 4.3 Slovakia 0.39
Sweden 4.2 Cyprus 0.53
Kosovo 3.3 Slovenia 0.81
Ireland 3.0 Luxembourg 0.84
Spain 3.0 Austria 0.96
Romania 2.8 Belarus 0.97
Belgium 2.6 Lithuania 1.01
France 2.5 Czech Republic 1.09
Bosnia and Herzegovina 2.4 Croatia 1.16

In the following graph, we are going to see the level of democracy in those two groups of countries to see if we can manage to retrieve useful information to understand whether democratic countries register higher or lower mortality rate compared to “autocratic” countries.

In this case, we can see a clear difference between democratic and non democratic countries. As we can see, both groups, high and low mortality rate can be observed in both countries regardless of their level of democracy. In the first one, we see that 7 countries out of 10 have a high democracy level. Therefore, in the second group (low mortality rate), this proportion is even higher; 8 countries out of 10 have a democracy level above 0.7.

So, to answer our first research question, we can say that there are not significant differences in mortality rates between democratic and non democratic countries. Although, countries with high level of liberal democracy are the one showing both highest and lowest mortality rate and, for this reason, we cannot define a relevant connection between the two variables.

Democracy and restrictions

In this section, we will focus on the relationship between countries’ level of stringency and their democracy degree. We will try to understand whether countries with a high level of democracy are setting lower pressure than the less democratic ones. In order to measure that, we will follow the same process as we did before, classifying countries in two groups, high and low level of liberal democracy.

Stringency index goes from 0 to 100, with 0 representing no restriction at all and 100 all restriction applied. In order to facilitate the understanding of our graph, we have classified the stringency index into five levels. Each level represents one range of stringency such that level 1 include values between 0 and 20, level 2 20 to 40 and so on. In this case, we want to identify which countries reached the last class (level 5). We added a dashed line to highlight the time frame for which those countries have been keeping such stringency and we reported the number of days on the side.

We notice, for our highly democratic group of countries, that only four of them have applied the highest level of stringency. These are “Belgium, Ireland, Portugal and Spain”. In the other groups, we have eight countries that achieved that level. These countries are “Albania, Belarus, Bosnia, Kosovo, Moldova, Romania, Russia, Serbia, Ukraine”.

As a result, we observe that not only less democratic countries have been more severe in term of applying restrictions rules but also, on average, they have lasted much longer than the more democratic ones.

However, it is still hard to assume a correlation between democracy and the level of stringency.They may have more strict rules because they have registered a higher number of cases than more democratic countries.

For this reason, in the next step, we will observe the evolution of the number of cases per 100.000 people through time by comparing to the evolution of stringency rules.

This graph provides us with exciting results. Firstly, it allows us to visualize the situation of covid-19 cases per 100.000 people in the two groups of countries. Since the highest levels of stringency were registered at the beginning of 2020, we will mostly focus on that period.

To begin with analysis, we start from the fact that in countries where the level of democracy is high, we have two different situations. Norway, Denmark, Estonia and Finland show a flat line and consequently have been less strict in terms of restrictions and, as it is shown, their level of stringency never went over 80 (lv.5 in this graph). On the other hand, Sweden and Switzerland have suffered an increase of cases in April, but still maintained a low level of stringency. As we have seen before, Sweden is now showing the highest mortality rate. This bring us to conclusion that their decision not to impose restrictions as the other countries did, was not a feasible one.

In contrast, we see that countries with a low level of democracy have been stricter in terms of restrictions and with the exception of Belarus and Hungary, they have all imposed severe policies to fight the pandemic. This is true even if the number of cases, in those countries, was not as high as in Spain, Portugal, Belgium, or Ireland.

Furthermore, it is clear that in this group of countries the high level of stringency seems to last longer, for an average of 66 days, compared to only 52 in the high democracy group.

These results drive us to a conclusion that less democratic countries are setting severer stringency levels that last longer than the ones in countries with stronger democracies.

We now want to verify if such conclusion can be applied in more detail for movement restrictions.

MOVEMENT RESTRICTIONS

STAY HOME

We will use the two clusters again, group 1 and group 2, where the first one represents countries with a high level of democracy, while the second one shows the low ones.

In the following graph, the colors show the strictness of the restriction “Stay-at-home”, with intensity from 0 to 3.

First of all, we notice that democratic countries never reached the highest level of restriction where people are not allowed to go out at all, with some minor exceptions. On the other hand, 40% of the authoritarian countries imposed such measures.

Also, we can see that Belgium, Portugal, Spain and Ireland have been quite shaky with applying such a policy, which means that they are frequently changing their strategy. At the same time, we can see similar behavior in Romania and Moldova, probably stemming from the fact that they are the only countries in group 2 to have a similar liberal democracy level. One reason for such fluctuation on restriction level is that democratic countries are hesitant in imposing the maximum stringency to stay at home, fearing the possibility of its citizens protesting.

However, we see that the rest of the countries in group 1, apart from Estonia, never reached level two either. This holds true even when they were facing a steep increase in the number of cases.

On the other hand, in group 2, we observe more stability on the restriction applied. The only countries who applied the maximum restriction for stay at home are Russia, Serbia, Bosnia and Kosovo. Looking at the cases graph, we see that the situation in those countries was not far from the average, and in term of mortality rate, they were the countries with the lowest numbers.

Overall, countries with a higher level of democracy seem to face more instability in applying the stay at home restriction, more precisely, they never managed to apply it at the full regime. In contrast, low democratic countries have been able to confine people at home during the peak of the emergency (March to June), even if indicators were showing a more favorable situation.

Table 13: Stay home restriction - Total restriction length in days and number of changes in restriction policy
country changes 0 1 2 3
Denmark 2 77 232 0 0
Estonia 2 259 0 50 0
Sweden 1 84 225 0 0
Switzerland 2 212 97 0 0
Norway 0 309 0 0 0
Belgium 7 181 5 123 0
Portugal 6 99 90 120 0
Spain 7 86 112 111 0
Finland 2 232 77 0 0
Ireland 7 128 74 107 0
country changes 0 1 2 3
Moldova 4 163 93 53 0
Romania 7 119 139 51 0
Albania 2 72 157 80 0
Kosovo 5 94 0 212 3
Hungary 4 126 131 52 0
Bosnia and Herzegovina 3 247 0 27 35
Ukraine 1 75 234 0 0
Serbia 5 97 149 3 60
Belarus 0 309 0 0 0
Russia 5 96 44 106 63

PUBLIC GATHERING

In both groups, it can be seen a reasonable level of stringency for public gathering restriction; however, this measure can be easier to apply since it does not affect the citizens’ will to move, but it impose a limit on social relations. This restriction shows some differences with other measures as countries started becoming very strict between March and May and, on average, the limit on public gathering stayed under 100 people almost everywhere until now, whereas, for other restrictions, we have observed that they have been eased during summer.

Sweden is the only democratic country which decided not to impose a limit below 10 people, while all the others did. Again, we have a similar behavior between Belgium, Spain and Portugal. Since March, the limit on public gathering has always been at his maximum in those countries, except during summer, when it has been raised to 10-100 people.

On the other hand, autocratic countries are showing a very similar pattern to the democratic ones. They have also imposed the highest level of measure. Only Hungary and Belarus stayed below the average. The latter, never imposed limitation on people movement, showing level 0 in both, stay at home and public gathering restriction.

For what we see, most of the countries, regardless of their level of democracy, have applied “Public Gathering Policy” since the beginning of the emergency. As a result, we can not observe any link between the level of democracy and public gathering policy.

HEALTH CARE POLICIES

Another question we are trying to answer is how different testing policies and contact tracing are between countries with different level of democracy.

We expect democratic countries to show an higher level of control over the spread of the virus.

TESTING POLICY

As we mentioned in the Introduction, several research papers state that highly democratic countries are better with dealing crisis than less democratic countries since the better governments treat their employers, citizens, the longer they will stays.

For this reason, we expect countries in group 1 (high democracy) to care more about people health, using testing and contact tracing activities in larger scale compared to countries with a lower level of democracy.

Looking at the graph, overall, we see that all countries followed an increasing trend, since the early stage of the pandemic. However, covid tests are rarely open to the public.

The majority of democratic countries allow people to get tested whenever they show symptoms. The only two countries who are now allowing open test are Denmark and Portugal while Switzerland did it only from June to September. We assume that they decided to do so because, during summer, many people were going on holiday and, to prevent an import of the virus from outside the countries, they allowed everyone to get tested.

On the other hand, less democratic countries follow a very similar trend as democratic ones. Kosovo is the only country which allows people to get tested only with symptoms and specific criteria. Another information that strikes us is Belarus. As we mentioned before, it never applied any movement restriction but, in term of the testing policy, together with Russia, Serbia and Ukraine, followed an open public testing policy. Belarus and Russia started using this strategy since the beginning of the pandemic while, Ukraine has slowly eased the criteria for testing over the year and reached open public test after summer.

Differently, Serbia faced a high increase in new cases from June to August and for this reason, they allowed everyone to get tested even without symptoms. Nevertheless, in July, they decided to stop, and they went back to level 1, where people are allowed to get tested only with symptoms and specific criteria, even though they were still registering an increase in new cases.

As for public gathering, we can not assume a positive relation between countries level of democracy and their testing policy. Therefore, we will continue with our analysis, interrogating our data about differences that might raise between democratic and non-democratic counties in term of contact tracing policy.

CONTACT TRACING

The first information we get from the graph is that countries with a democracy level lower than 0.4 (Bosnia excluded), started on average, to apply contact tracing, for all cases, since April. In this group, the only countries who did not are Bosnia, Kosovo and Romania. The first never practiced such activity while the other two have been doing it only for some cases. Finally, Romania started tracing back the infection recently.

In contrast, in the first group, we see a dominance of level 1 over the whole period. We immediately thought about the difficulty incurred in countries where population density is high. Indeed, it is harder to trace back someone’s movements in such countries. At the same time, looking at the population density for those countries, we see that Belgium and Switzerland are the ones with the highest; however, they are now applying contact tracing for all cases.

Finally, we also see that low democratic countries show more stability in this activity, and they never downgraded their strategy. On the other hand, similarly to stay at home restriction, countries in group 1 are showing more variability on whether doing the contact tracing for all cases or only for some of them.

Overall, we can affirm that countries with a low level of democracy are more inclined to engage in contact tracing. This result appears to be in contrast with our initial expectations.

Conclusion

No country would have ever expected this situation to last that long. After one year, we can say that many countries are sharing similar responses to the emergency, applying similar restriction aimed to reduce the spread of the virus. Therefore, with this paper, we aimed to provide a definite answer about whether democracy is the best regime to deal with crises.

In Europe, we observed that countries with a higher level of democracy are less strict in applying restrictions. We came up with our assumption: in countries where democracy level is high, it is more difficult to apply restrictions because their citizens are used to be more involved in governing the country. However, the authorities, being in a state of emergency, need to act promptly with restrictions that limit people freedom. As those decisions did not come from the people, even if it is for the people, it creates popular discontent.

We started analyzing the difference in the mortality rate in Europe. We saw that democratic countries have been having higher rates of fatality but, at the same time, the ones showing the lowest rates are included in the high democracy group too. Hence, we have not seen any correlation between government regime and mortality rate.

Secondly, we decided to check which of the two groups of countries have been, and are, more strict in term of movement restrictions. Low democratic regions seem to be more stable when applying such measures, confirming our theory that in democratic countries, it is harder to limit people’s freedom to move. On the other hand, for public gathering, the two clusters have shown very similar results.

We moved then to the analysis of two health policies applied by all the countries analyzed. Low democratic countries are more strict in term of contact tracing. Despite that, through the analysis of testing policies, we can confirm the main findings of the paper called “Democracy, Authoritarianism, and COVID-19 Pandemic Management: The Case of SARS-CoV-2 Testing” written by German Petersen. It is observed no clear relationship between countries democracy level and their testing policy.

Limitation

To improve the quality of our research, we should increase the sample of countries. However, the main reason we focused only on Europe countries is that they show a very similar pattern in many aspects. That allowed us to evaluate countries stringency measures and its level of democracy, making the famous assumption “ceteris paribus”, meaning “all else being equal”. Our results hold only when this hypothesis is highlighted. However, we are aware that this assumption is extreme. In order to ensure reliable results, the analysis should include more variables.

Future work

High democracy level does not mean that citizens have full trust in the government. According to the article published by “CEIP”, a global think tank, Many European democracies have faced declining trust since the economic crisis that took place in 2008. Spain, France, Italy and the UK are ranked particularly low. Furthermore, it is suggested that governments with high levels of trust can effectively maintain lockdowns. Trust enabled some countries to convince their citizens to allow mass testing and quarantine before the virus’s effects were widely seen, allowing them to stop the spread early.

As future work, we will test the validity of this argument using corruption data, in order to identify the level of trust people have towards their representatives. The next question we will try to answer could be: “Do highly trusted governments face less difficulties in applying restrictions compared to less trusted ones?”